A comparison of five models in predicting surface dead fine fuel moisture content of typical forests in Northeast China

Abstract

IntroductionThe spread and development of wildfires are deeply affected by the fine fuel moisture content (FFMC), which is a key factor in fire risk assessment. At present, there are many new prediction methods based on machine learning, but few people pay attention to their comparison with traditional models, which leads to some limitations in the application of machine learning in predicting FFMC.MethodsTherefore, we made long-term field observations of surface dead FFMC by half-hour time steps of four typical forests in Northeast China, analyzed the dynamic change in FFMC and its driving factors. Five different prediction models were built, and their performances were compared.ResultsBy and large, our results showed that the semi-physical models (Nelson method, MAE from 0.566 to 1.332; Simard method, MAE from 0.457 to 1.250) perform best, the machine learning models (Random Forest model, MAE from 1.666 to 1.933; generalized additive model, MAE from 2.534 to 4.485) perform slightly worse, and the Linear regression model (MAE from 2.798 to 5.048) performs worst.DiscussionThe Simard method, Nelson method and Random Forest model showed great performance, their MAE and RMSE are almost all less than 2%. In addition, it also suggested that machine learning models can also accurately predict FFMC, and they have great potential because it can introduce new variables and data in future to continuously develop. This study provides a basis for the selection and development of FFMC prediction in the future.

Comparative Analysis of Machine Learning-Based Predictive Models for Fine Dead Fuel Moisture of Subtropical Forest in China

“…The data are collected by the acquisition host, while the fuel loading and meteorological data can be obtained through the Android mobile phone app or Beidou satellite transmission. The entire device is powered by an accumulator and a solar panel to ensure continuous operation of the monitoring device [37].…”

“…The principle is that after introducing all independent variables into the equation, according to the degree of influence of independent variables on dependent variables, a correlation significance test coefficient is used to eliminate the independent variables without significance. The independent variables with statistical significance are selected, and regression models are established to reflect the correlation between multiple independent variables and dependent variables [14,15]. Stepwise linear regression is based on the 'MASS' package in the R-project.…”

“…But remote sensing has limited observation ability under the forest canopy, so its prediction accuracy is low [12,13]. The meteorological element regression method is simple in design, but it is limited by the physical and chemical properties of the fuel itself and the study area, and the spatio-temporal extrapolation effect is usually poor [14,15]. The process model is based on physical descriptions, which is difficult to apply in practice because of its complexity [16,17].…”

“…Such methods have found extensive recent application in many fields, including fuel moisture content prediction [18,19]. Fan et al analyzed the dynamic changes and driving factors of typical dead surface fuel moisture content in the cold temperate zone of northeast China and compared the performance of machine learning algorithms with traditional equilibrium moisture content prediction models [14]. Miller et al estimated the surface fuel moisture content of the continental United States by establishing a temporal convolutional network [20].…”

“…The bulk of previous research on fuel moisture content focuses on cold temperate coniferous forests and Mediterranean climate forest areas [14,15,29]. The subtropical zone, which accounts for 11% of the global forest area [30], is one of hardest hit by forest fires [31].…”